Dispersion of Ferrofluid Aggregates in Steady Flows: Related Articles

Dispersion of ferrofluid aggregates in steady flows
Alicia M. Williams and Pavlos P. Vlachos Citation: Phys. Fluids 23, 127102 (2011); doi: 10.1063/1.3670012 View online: http://dx.doi.org/10.1063/1.3670012 View Table of Contents: http://pof.aip.org/resource/1/PHFLE6/v23/i12 Published by the American Institute of Physics.
Related Articles
Fluid flows in a librating cylinder Phys. Fluids 24, 026603 (2012) Dipole evolution in rotating two-dimensional flow with bottom friction Phys. Fluids 24, 026602 (2012) Low-dimensional model of spatial shear layers Phys. Fluids 24, 014108 (2012) Three-dimensional swirling flows in a tall cylinder driven by a rotating endwall Phys. Fluids 24, 014101 (2012) Three-dimensional evolution of flow structures in transitional circular and chevron jets Phys. Fluids 23, 124104 (2011)
Additional information on Phys. Fluids

Journal Homepage: http://pof.aip.org/ Journal Information: http://pof.aip.org/about/about_the_journal Top downloads: http://pof.aip.org/features/most_downloaded Information for Authors: http://pof.aip.org/authors
Downloaded 23 Feb 2012 to 200.59.229.134. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
PHYSICS OF FLUIDS 23, 127102 (2011)

Alicia M. Williams1 and Pavlos P. Vlachos2
Lawrence Livermore National Laboratory, Livermore, California 94550, USA Mechanical Engineering Department, Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061, USA
2 1
(Received 8 April 2011; accepted 7 November 2011; published online 14 December 2011) Using focused shadowgraphs, we investigate steady ows of a magnetically non-susceptible uid interacting with ferrouid aggregates comprised of superparamagnetic nanoparticles. The ferrouid aggregate is retained at a specic site within the ow channel using two different applied magnetic elds. The bulk ow induces shear stresses on the aggregate, which give rise to the development of interfacial disturbances, leading to Kelvin-Helmholtz (K-H) instabilities and shedding of ferrouid structures. Herein, the effects of bulk Reynolds number, ranging from 100 to 1000, and maximum applied magnetic elds of 1.2 105 and 2.4 105 A/m are investigated in the context of their impact on dispersion or removal of material from the core aggregate. The aggregate interaction with steady bulk ow reveals three regimes of aggregate dynamics over the span of Reynolds numbers studied: stable, transitional, and shedding. The rst regime is characterized by slight aggregate stretching for low Reynolds numbers, with full aggregate retention. As the Reynolds number increases, the aggregate is in-transition between stable and shedding states. This second regime is characterized by signicant initial stretching that gives way to small amplitude Kelvin-Helmholtz waves. Higher Reynolds numbers result in ferrouid shedding, with Strouhal numbers initially between 0.2 and 0.3, wherein large vortical structures are shed from the main aggregate accompanied by precipitous decay of the accumulated ferrouid aggregate. These behaviors are apparent for both magnetic eld strengths, although the transitional Reynolds numbers are different between the cases, as are the characteristic shedding frequencies relative to the same Reynolds number. In the nal step of this study, relevant parameters were extracted from the time series dispersion data to comprehensively quantify aggregate mechanics. The aggregate half-life is found to decrease as a function of the Reynolds number following a power law curve and can be scaled for different magnetic elds using the magnetic induction at the inner wall of the vessel. In addition, the decay rate of the ferrouid is shown to be proportional to the wall shear rate. Finally, a dimensionless parameter, which scales the inertia-driven ow pressures, relative to the applied magnetic pressures, reveals a power law decay C relationship with respect to the incident bulk ow. V 2011 American Institute of Physics. [doi:10.1063/1.3670012]
I. INTRODUCTION
Ferrouids are stable suspensions of subdomain particles that become strongly magnetized in the presence of a magnetic eld. The interaction of ferrouids with magnetic elds adds a dimension of physics beyond those of conventional uid mechanics and is the basis of the eld of ferrohydrodynamics. Studies of ferrouids include those focusing on the apparent change in viscosity of a ferrouid under a magnetic eld1 or on the behavior of these magnetically susceptible uids due to changes in magnetic eld strength.2 Moreover, uid instabilities have been the subject of signicant scientic works, including the spike-shaped protrusions from a ferrouid free surface known as the normal eld instability,3,4 the labyrinthine instability,5 and the Rayleigh-Taylor instability.6 The Kelvin-Helmholtz instability in ferrouids has also been studied in several papers analytically,79 but has been largely neglected experimentally. Ferrouids have been at the center of many new inventions since the advent of ferrouid research in the 1960s, including improved dampers and seals.10 Some of the newest
1070-6631/2011/23(12)/127102/11/$30.00
ferrouid applications have been focused on their integration in a new technique for targeting disease treatment in the human body.1115 Magnetic drug targeting (MDT) involves magnetically guiding droplets of a biocompatible ferrouid bonded reversibly to specic chemotherapeutic agents, which are injected into the blood or lymphatic circulation system.15 Some trials performed on lab animals have shown that 50% of the typical systemic dose applied to treat cancerous tumors can result in remission of the tumor after only one treatment of anticancer drugs with MDT.11,1517 Though these studies have provided proof that the MDT application has promise as a clinical modality, only a few physics-based works have examined the interaction of the ferrouid aggregate with an incident non-magnetic ow. Understanding the governing physics and scaling of the ferrouid dispersion and retention could aid in the development of improved MDT methodologies. Initial experimental contributions towards understanding ferrouid-uid interactions were published by Ruuge and Rusetski18,19 as well as Ganguly et al.2022 In the work of Ruuge and Rusetski, an aggregate of ferrouid was
C V 2011 American Institute of Physics
23, 127102-1
127102-2
A. M. Williams and P. P. Vlachos
Phys. Fluids 23, 127102 (2011) TABLE I. Experimental test matrix to study aggregate dispersion. Applied magnetic field, A/m 1.2 10 2.4 105
5
accumulated in a tube prior to the startup of bulk ow. Ferrouid dispersion was measured using gamma-active tagging over a wide range of ows, and the authors examined the long-time dispersion of the ferrouid. In Gangulys work, shadowgraphs were used as a means to analyze the behavior of ferrouid aggregates as ferrouid is injected into a test section. One of the primary results from these initial studies is the periodic buildup and shedding of a ferrouid aggregate accumulated by a permanent magnet at Reynolds numbers below 400 observed over several minutes of data acquisition. The results of these works will be discussed in further detail in the context of the results obtained from present experimentation. Herein, we extend upon previous works by examining the interaction of a ferrouid aggregate with a non-magnetic uid under steady ows for bulk ow Reynolds numbers (Re) ranging between 100 and 1000 and under the action of two separate applied magnetic elds. In order to extract the dynamics of the ow-ferrouid interaction, measurements are performed using focused shadowgraphs and the recorded images are quantitatively analyzed using digital image processing. Under the action of the magnetic eld, the aggregate behaves similarly to a bluff-body, inducing shear-layers and generating a wake as well as vortex shedding. However, since the aggregate bluff-body is essentially comprised of a uid, it deforms dynamically and according to the shear and normal stresses induced by the bulk ow and the magnetic eld. The result is reduction in size of the body as the induced forces cause the ferrouid to break up and disperse downstream, in the direction of the bulk ow. This paper will rst present the experimental method for recording focused shadowgraphs as well as the algorithms for quantitatively post-processing the raw frames. The aggregate dynamics will be presented in terms of the evolution of the aggregate size and shedding frequency over time as functions of both the magnetic eld and Reynolds number. Finally, single parameters, including a dimensionless parameter, are used to delineate aggregate behavior and are presented as a means to capture the aggregate physics observed across the identied regimes.
II. EXPERIMENTAL METHODS
Reynolds numbers investigated 100 to 1000 in steps of 100 100 to 600 in steps of 100, 900, and 1000
The data were obtained using focused shadowgraphs, which are particularly effective when investigating phenomena that involve uids with distinct density gradients or an opacity differential that generates a high contrast interface, as is the case of ferrouid and water studied herein. Illumination for the shadowgraph background was provided by a 250 Watt halogen white light source that was directed through spherical and ground glass lenses to illuminate a region of interest with uniform intensity of 2 by 15 cm in size. The lighting was directed at the square channel, illuminating the region in which the ferrouid was retained. A schematic depicting the setup for the shadowgraph experiments is shown in Figure 1. A digital charge-coupled device (CCD) camera sampling at 50 Hz was used to image the side plane of the test section where the ferrouid aggregate was retained. A total of 3000 images were acquired for each case, corresponding to a recording time of 60 s. The image magnication was 63 microns/pixel with an image size of 1020 by 500 pixels. This interrogation region encompassed the channel height as well as 4 Dh in the streamwise direction. This experiment compares differently to those of Ganguly et al. in that the focus was placed on the acute transient behavior of the aggregate and its interfacial dynamics, whereas previous works focused on long-term residence time of ferrouid aggregates at lower Re. In each case, a 2 ml volume of water-based ferrouid (EMG 705, Ferrotec, Inc.) was injected 4 Dh upstream of the
For the experiment presented here, ferrouid is introduced and is allowed to interact with a steady ow in a square channel with hydraulic diameter Dh 0.015 m. A square channel was selected over a circular channel to avoid introducing additional three-dimensional effects to the dynamics of the aggregate, focusing on the basic aggregate physics. The Reynolds numbers based on the bulk ow parameters (average velocity Uo, hydraulic diameter Dh, density of water q, dynamic viscosity of water l, and kinematic viscosity of water ) ranged between 100 and 1000. In addition, two magnetic elds with values of 1.2 and 2.4 105 A/m were considered in order to allow for investigating the scaling of the magnetic forces. The experimental test matrix is provided in Table I. For the second (higher) magnetic eld, Reynolds numbers of 700 and 800 were later deemed unusable and were not further analyzed.
FIG. 1. (Color online) Schematic of focused shadowgraph experiment. A high-speed digital camera imaged the side prole of the ferrouid aggregate.
127102-3
Phys. Fluids 23, 127102 (2011)
region of interest before each test in quiescent conditions at a rate of 1 ml per min. EMG 705 ferrouid has a saturation magnetization of 22 mT, a nominal density of 1.19 g/cc, and a nominal magnetic particle concentration of 3.9% by volume, with a mean particle diameter of 10 nm. The ferrouid was diluted to 2% of its original concentration by mixing with water. Injection was facilitated by a needle aligned parallel to the test section walls and mounted ush to the bottom wall of the channel. A KD Scientic 230 syringe pump was utilized for all injection, with an estimated accuracy of 1% full-scale volumetric ow. The ferrouid was captured and retained by a 0.5 Tesla permanent magnet that was obliquely oriented with one of its vertices closest to the test section, resulting in a maximum magnetic eld of 1.2 105 and 2.4 105 A/m at the inner test section wall. The magnetic eld was varied by means of a micro-positioning traverse. The bulk ow (water) was switched on after the ferrouid was accumulated and settled, owing in the positive x-direction. The ow rate was measured using a Transonic Systems T110 ultrasonic ow meter with an absolute accuracy of 10%. The test section was 1 meter in length to enable fully developed ow for all Reynolds numbers examined in this work relative to the quiescent ow aggregate height of 7 mm. Flow was ramped up over 0.1 s, which ensured quasisteady acceleration for all cases based on the inertial time scale t=U2 . It should be noted that the aggregate formation o and subsequent interactions are only dependent on the presence of the magnetic eld. When the magnet is removed, the aggregate disperses instantaneously in the positive x-direction as the ferrouid is superparamagnetic, and does not retain its magnetism in the absence of a magnetic eld. The magnetic eld was characterized using a DC Gaussmeter (Alphalab, Inc., 1 Gauss resolution at 20 000 Gauss dynamic range). The applied magnetic eld is primarily twodimensional, due to the oblique orientation of the magnet. The oblique orientation also provides high magnetic eld gradients, with the primary direction of the eld being perpendicular to the direction of bulk ow. Contours of the spatial distribution of the resulting normal magnetic induction are shown in Figure 2, with out-of-plane magnetic induction neglected in subsequent analysis. This distribution is used to estimate the magnetic eld strength at the corresponding ferrouid-uid interface heights for analysis presented in proceeding sections. Alterations of the magnetic eld due to the presence of the magnetized ferrouid were not considered in this work. The applied magnetic eld is calculated from the measured magnetic induction using the formula H B ; l0 (1)
FIG. 2. Contours of magnetic eld induction in the z-direction (Tesla). The highest eld gradients are near the magnet tip at (x, y) (0, 0). The magnetic eld is shown relative to the channel bottom for both maximum magnetic elds of 0.30 and 0.15 Tesla.
frequency of the Kelvin-Helmholtz rolls shed from the aggregate. First, the methodology used to extract the aggregate size in time will be discussed, and then, a separate process used to determine the shedding frequency over time will be presented. Figure 3 shows the steps taken to measure the evolving size of the aggregate over time. The interface between the ferrouid and water was sampled for each snapshot of the ow using software written in-house in conjunction with the MATLAB image processing toolbox. The grayscale images were converted to black and white using Otsus23 grayscale thresholding algorithm. Otsus method assumes uniform image illumination and a bimodal image histogram, used as a basis to binarize each image. Regions of connected pixels in the binary images were identied and sorted by size for each frame, which represent the ferrouid aggregate or vortex shedding. The size of the aggregate was obtained using this method by selecting the largest region, which was conrmed through the output of the binarized image. The aggregate size, measured from the binary images, was corrected
where lo is the permeability in a vacuum (4p 107 N/A2) and B is the magnetic eld induction. In this work, the maximum value of H, or Hmax, is used to denote the value of the maximum magnetic eld at the location of the inside wall the ow channel. The raw frames obtained by the focused shadowgraph technique were post-processed to determine the aggregate size evolution in time as well as to detect the shedding
FIG. 3. The image processing used to transform a grayscale raw image (a) to a binary image (b) to an image in which the instantaneous aggregate was isolated (c).
127102-4
Phys. Fluids 23, 127102 (2011)
by weighting with the grayscale intensity of the pixels in the region of the aggregate. The intensities account for the local ferrouid concentration and correct the aggregate size by up to 10%. Pixel intensities are a factor when calculating the time-rate of dispersion, particularly as the aggregate deforms in the x and y-directions. Image binarization was performed based on the intensity distribution of each individual image, and, therefore, does not account for bulk ow-driven reductions in the local ferrouid concentration. The variation in initial aggregate size was estimated to 5% across all cases. In addition, the random error in the aggregate size measurement is estimated to be /0.5%, based on the measured variance between two successive frames of a ferrouid aggregate retained in quiescent ow. In the case in which the aggregate disperses through the unsteady shedding of vortices, it was desired to measure the shedding frequency over the data acquisition time. Similarly to the method used to isolate the aggregate size, the images were binarized, focusing on the region just downstream of the aggregate. Figure 4 shows a sample image with the region of interrogation, which is subdivided into ve smaller windows for the binarization that have 15 pixel width by 200 pixel height. Tall windows spanning the majority of the channel height are required to consistently capture the shed ferrouid as the height of the aggregate changes signicantly over time in the case of K-H shedding. The fraction of black pixels relative to the total number of pixels per window was recorded for each of the ve windows over time, and then the average fraction from the individual windows was calculated for each frame, corresponding to a single time instant. The measured fraction of dark pixels with respect to the total number of pixels in the window is shown in Figure 5 for a sample case with a bulk Reynolds number of 500 and magnetic eld of Hmax 1.2 105. The time series is detrended for processing as shown in Figure 5(b). After detrending, the signal remains nonstationary and nonergodic, and, therefore, for the spectral analysis, we used short time Fourier transform (STFT).24,25 A sample STFT corresponding to the data from Figure 5 is shown in Figure 6, where 512 point Hanning windows were used in conjunction with a 75% overlap for a frequency resolution of 0.1 Hz. The shedding frequencies were extracted from each band in the STFT by determining the frequency corresponding to the peak value from the STFT corresponding to each time band as in Figure 7. These shedding frequencies were conrmed by
FIG. 5. (a) Sample time series of raw detected fraction of ferrouid passing through image windows through time. (b) De-trended fraction of ferrouid passing through image windows over time.
manually estimating the shedding frequency using the raw frames. This provides information complementary to the aggregate dispersion measurements in that the shedding frequencies indicate rates of aggregate decay.
FIG. 4. Binarization of ferrouid shed from aggregate to determine shedding frequency.
FIG. 6. (Color online) Sample short time Fourier transform for Re 500 with Hmax 1.2 105 A/m.
127102-5
Phys. Fluids 23, 127102 (2011)
FIG. 7. Extraction of amplitude versus frequency during one time period extracted from the STFT in Figure 6. The peak energy corresponds to the shedding frequency of the aggregate.
III. RESULTS A. Aggregate dispersion characteristics
The large-scale changes of the aggregate are characterized by three primary dispersion regimes: stable, transitional, and shedding. A representative set of shadowgraph images of cases with Reynolds numbers corresponding to the three regimes is shown in Figure 8. For Reynolds numbers below 400, slight aggregate stretching occurs in the positive x and y-directions, and the aggregate is displaced downstream. No measurable reduction in aggregate mass is observed in two dimensions as interrogated by side-oriented digital imaging in these cases. The aggregate is three-dimensional, however, during the largely static behavior of these low Re cases, aggregate dynamics are largely two-dimensional. For a Reynolds number of 400, stretching in the positive x and y-directions occurs initially, resulting in a slight apparent increase in aggregate size due the two-dimensional measurement method. The biaxial stretching ultimately overcomes the magnetic body forces, since the aggregate expands beyond regions of high magnetic eld gradients wherein, it is controlled by the magnet. Stretching in the direction of the bulk ow leads to the growth of interfacial waves on
the aggregate and slight wisps of ferrouid are shed. As the Reynolds number is increased to 500 or greater, the aggregate centroid translates signicantly in the positive x-direction when exposed to ow pressure, which also results in signicant aggregate stretching along this axis. As a result of both the aggregate deformation and interfacial excitation, the Kelvin-Helmholtz instability develops and leads to shear layer shedding. As a result, the aggregate decays rapidly as ferrouid vortices are shed, with threedimensional structures apparent in the shadowgraph data. The primary difference between the two eld inductions becomes apparent at Re 300 as shown in Figure 9. The case with the lower magnetic eld remains in the stable regime, while at Hmax of 2.4 105, the aggregate destabilizes and washes away. As the Reynolds number increases past 400, the effect of the increased eld is diminished in terms of observable differences in regime-based dynamics, as inertial effects dominate the aggregate characteristics. The fundamental mechanics of aggregate behavior are illustrated in Figure 10. Three time instants of the main aggregate outline are plotted for a Reynolds number of 500, Hmax of 2.4 105, beginning with the primary ferrouid-uid boundary in quiescent conditions. As ow is ramped, the centroid of the aggregate translates in the positive x-direction, while the aggregate stretches out as ow is ramped. The aggregate then rebounds towards its starting location (centered over the magnet tip), which fosters interfacial perturbations and the initial appearance of the K-H instability. For all Reynolds numbers, the relative size of the aggregate (a/ai) is plotted versus time in Figure 11 for maximum elds of 1.2 105 and 2.4 105 A/m. The dispersion was calculated as the normalized sum of the number of pixels in the edge-detected aggregate with respect to the initial aggregate size, ai. The normalization using the initial aggregate size provides the relative change in aggregate size over time. When the Reynolds number is lower than 400 with Hmax of 1.2 105, an increase of the apparent aggregate size is manifested by a slightly upward sloping line, indicating aggregate stretching due to shear stresses. The stretching increases from 3.8 to 6.3, and nally, 8% for Reynolds numbers between 100 and 300. The initial aggregate stretching is similar when Hmax is 2.4 105 for Reynolds numbers up to 300. At the transition point of Re 400 with Hmax 1.2 105 and Re 300 with Hmax 2.4 105 A/m, a conspicuous
FIG. 8. Shadowgraph images for three time instants of 0, 25, and 50 s with Reynolds numbers of 200, 400, and 600. For a Reynolds number of 600, Kelvin-Helmholtz shedding occurs, causing the aggregate to wash away.
127102-6
Phys. Fluids 23, 127102 (2011)
FIG. 9. The impact of magnetic eld strength for magnetic elds of 1.2 105 and 2.4 105. For the stronger magnetic eld, the aggregate destabilizes and K-H shedding occurs from the aggregate approximately 25 s after the start of ow.
increase in the aggregate size occurs initially, stretching up to 10% and 7%, respectively. After stretching, wisps of ferrouid are periodically shed from the aggregate, and the aggregate begins to disperse. The incipient destabilization at transition for the higher magnetic eld is more pronounced, and therefore, signicant aggregate dispersion occurs, resulting in a size reduction of 14% of the initial aggregate. As the Reynolds number is increased past the transition point for both eld strengths, the initial stretching of the aggregate occurs rapidly, within the rst seconds after ow ramp-up. After the initial stretching, the aggregate destabilizes and decays exponentially through Kelvin-Helmholtz shedding. The evolution of shedding of the aggregate is captured in Figure 12, where the Strouhal number (St) is plotted against the normalized specic volume-based magnetic body force, FM, for the two applied magnetic elds. The Strouhal number is calculated based on the shedding frequency of the ferrouid aggregates St fs h ; Ub (2)
the interface between the ferrouid aggregate and the bulk ow using the aggregate height extracted from the imaging data. A value for FM in terms of force per unit volume was calculated at each time instance and is normalized to the magnetic body force measured at the start of each case, FM,0. This normalization delineates how the aggregate size and body forces change relative to the Strouhal number plotted on the y-axis. Shedding frequencies were measured from the raw frames as described in the methods section, with the rst 5 s
where Ub represents the blockage velocity of the water owing over the aggregate, which is a function of aggregate height. The parameter h represents the instantaneous aggregate height. fs is the measured shedding frequency of the ferrouid whorls in radians per second. The magnetic forces are captured by the specic volume-based magnetic body force, FM Fm MDH; (3)
which is calculated using the magnetization of the ferrouid, M (Tesla), and the spatially resolved applied magnetic eld, H (Amperes/meter). The magnetization was determined using Langevins equation, and body force was calculated at
FIG. 10. Line plot illustrating transition of the aggregate into a K-H shedding regime. (a) Quiescent ow, t 0 s, (b), aggregate stretching, t 0.56 s, and (c) main aggregate outline during ferrouid shedding, t 9 s.
FIG. 11. Plots of the apparent aggregate size versus time with Hmax of 1.2 105 and 2.4 105 A/m.
127102-7
Phys. Fluids 23, 127102 (2011)
of data acquisition neglected since large plugs of ferrouid initially slough off from the aggregate as ow is ramped up. Flow startup is distinguished from the startup of shedding, though this large perturbation of the aggregate appears to initiate K-H shedding. For post-processing purposes, the height of the aggregate was downsampled with 512 point subsets of the signal and averaged with 75% overlap, therefore, matching the temporal resolution of the shedding frequencies. Many of the dynamics observed in the shadowgraphs are quantied in Figure 12. Application of a higher magnetic eld results in higher shedding frequencies initially compared to lower magnetic elds. Moreover, the shedding frequencies are highest at the onset of shedding, when the aggregate is the largest and subjected to the greatest shear from the ow. The high shedding frequencies at the start of each case also drive the initial rapid drop in aggregate size, as shedding is the mechanism by which the aggregate disperses. From Figure 12, initial values of St start at approximately 0.2 and increase as Re increases. Roshko26 and Williamson27 reported St varying between approximately 0.16 and 0.21 for Re varying between 100 and 1000. Using these St to predict the initial shedding frequency based on the blockage velocity associated with the initial aggregate height after ow ramp up (transition to shedding), the calculated values range from 0.6 Hz for Re 100 to
8 Hz for Re 1000. These shedding frequencies are close to the measured values for cases where K-H shedding was observed, but also assume a fully rigid interface. Measured shedding frequencies (and resulting St) in Figure 12 are generally higher than the prediction, which is attributed to the dynamic interaction of the ferrouid interface with the bulk ow. However, for low Re, no shedding was measured from the ferrouid aggregate. In Figure 12, the normalized value of FM is also analogous to time, since the magnetic body forces at the aggregate-ow interface increase as the aggregate washes away. As wash away occurs, the aggregate height reduction acts to reduce the shedding frequency due to decreased shear stresses from the incident ow. Conversely, the magnetic body forces are increasing and assist in aggregate retention, further resisting shedding. Unlike cases with signicant dispersion, low Reynolds numbers that do not exhibit any shedding behavior have zero-valued St in Figure 12, with relatively stable values of FM/FM,0. On the other hand, cases that are transitioning to shedding have increasing Strouhal numbers over time, and very slight amounts of aggregate decay. These aggregates increase in height in the y-direction initially due to stretching, and then decrease as the aggregate begins to disperse. The height decrease occurs as the shedding frequency increases and ferrouid advects away from the aggregate. Cases with large amounts of shedding have magnetic body forces that increase consistently over time. For transitional cases, this data set captures only the increase in St as the aggregate initially transitions into a shedding mode. If additional temporal data were available, we would anticipate that the St will eventually begin to decrease as the aggregate height drops subsequent to ferrouid vortex shedding, following curves similar to those shown at moderate Reynolds numbers. Additional insight into the relationship between aggregate shedding and aggregate size is provided by examining the aspect ratio of the aggregate. Figure 13 shows the aspect ratio measured at the ferrouid aggregate half-life, as dened by the aggregate length divided by twice the aggregate height. Consistent increases in the aspect ratio occur as a
FIG. 12. Strouhal number versus normalized magnetic body forces for (a) Hmax of 1.2 105 and (b) Hmax of 2.4 105 A/m. Low Re cases have values of f/fo near 1 and have a continuously zero value of St due to lack of shedding phenomena.
FIG. 13. Plot of the aspect ratio calculated at the time corresponding to the aggregate half-life as a function of Reynolds number.
127102-8
Phys. Fluids 23, 127102 (2011)
function of Reynolds number, indicating relative aggregate lengthening in the streamwise direction due to ow shearing. Independently, examining the aggregate length, a slight decay occurs as a function of Re. However, the height of the aggregate at the half life reduces by a factor of approximately three between low Re and high Re. The aspect ratio data is also useful for relating to generalized uid mechanics studies, which provide additional data revealing the interaction between the ferrouid aggregate and bulk ow. In an analytical study of compliant bodies subjected to moderate cross-ows, Yokoyama and Mochizuki28 found that lower stiffness compliant cylinders subjected to cross ows deform and elongate into ovular shapes as they shed vortices. Paired with this observation was an increase in drag coefcient as the compliant cylinder lengthened, and as the stiffness of their cylinder was increased at a Re of 400. Higher drag forces imply that greater magnetic body forces would be required to maintain the aggregate as it lengthens, even neglecting the limited range of inuence of the applied magnetic eld. The change in drag coefcient found by Yokoyama and Mochizuki resulting from aspect ratio changes elucidate the signicantly different aggregate decays at moderate Reynolds numbers between 400 and 700, where the nal aggregate sizes diminish disproportionately between the fully stable cases and cases where shedding and aggregate washaway occurs. Yokoyama and Mochizuki additionally showed that not only do stiffer cylinders have higher drag coefcients, higher shedding frequencies are measured. As observed in Figure 12, our results are consistent with their ndings for compliant cylinders, though the source of the stiffness change was from a magnetic eld, rather than from the generalized mass-spring-damper system that they investigated. Figure 12 also shows relative aggregate lengthening as the Re is increased, and therefore, one can extrapolate that the aggregate is also subjected to higher drag coefcients as a result of the changing aspect ratio.
B. Aggregate decay scaling
FIG. 14. Plot of the time to reach 50% of the original aggregate size as a function of the Reynolds number.
Thus far, the temporal evolution of the aggregate under different ows and magnetic elds has been presented. However, the aggregate behavior can be more completely represented by distilling the time-series data into cumulative physical parameters. Therefore, we will now investigate the aggregate physics using the aggregate half-life, wall shear rate, and a dimensionless parameter developed for the ferrouid-shear ow system. First, we examine the half-lives of dispersing ferrouid aggregates. For cases where at least a 50% drop in the aggregate size was recorded, the corresponding times are plotted as a function of the bulk ow Reynolds number in Figure 14. This data shows that elevated magnetic elds increase the rate of decay, resulting in faster washaway of the aggregate down to 50% size. This nding is complementary to Figure 12, which shows that the Strouhal number increases when the applied eld is elevated. Generally, increasing the magnetic eld results in aggregate stiffening due to increased magnetization of the aggregate, in the case where the aggregate is magnetized below the saturation magnetization value
of the ferrouid.29 In this experiment, the aggregate magnetization is on the order of 105 Tesla at the aggregate interface, signicantly below the saturation level of 0.022 Tesla. In a system containing a ferrouid interacting in a shear ow, the magnetic forces oriented normal to the bulk ow can augment both the surface tension and body force effects associated with the development of the Kelvin-Helmholtz instability. This phenomenon is well documented, though largely through analytical studies.9,30 Therefore, result of the increased magnetic eld is faster aggregate dispersion, and lower half-life times. In addition, Figure 14 shows that the half life time is inversely dependent on the Reynolds number. The Reynolds number dependence on half life in a ferrouid system was also investigated by Ganguly et al.22 Their experimental system examined ferrouid accumulation and washaway an incident pulsatile ow with a 1 Hz frequency. They proposed an empirical power-law curve relating aggregate half-life (t50%, time at which the aggregate decays to half of its initial size) to the bulk ow Reynolds number by a power of 2.2. Figure 14 includes the power-law based t of the projected half-life, as well as the half-life times determined for ferrouid decay in this work. Symbols corresponding to data collected in the present experiment are plotted in contrast to the original curve t from Ganguly et al. It is apparent that the power law curve from Ganguly et al., though a good t for their dataset, does not capture the full range of ferrouid aggregate dynamics. The discrepancy is expected, given that Ganguly et al. were examining a pulsatile ow, where ferrouid was also accumulating at the magnet under an incident ow. Moreover, the curve from Ganguly et al. was generated from data where a single magnetic eld strength and gradient were used, with an induction of 0.28 Tesla at the inner edge of the ow channel. However, the curve developed from their data remains relevant. Using the same exponent as the Ganguly curve t, the predicted 50% time, scan be corrected so that it is representative of the
127102-9
Phys. Fluids 23, 127102 (2011)
data collected in the present work by dividing a constant of 107 by the maximum value of the magnetic eld at the inner edge of the ow channel each set of cases. This magnetic eld is representative of differences in the peak body forces applied to the ferrouid. The result of this correction is depicted in Eq. (4). The results of the curve t are plotted in Figure 14 with dashed and solid lines for Hmax of 2.4 105 and Hmax of 1.2 105 A/m, respectively. t50% 107 Re2:2 Bmax (4)
After the time to decay to half of the original aggregate size, we will next consider the rate of change in aggregate size. The time-averaged rate of change of the aggregate as a function of the wall shear rate is shown in Figure 15. The rate of change was calculated using a second order, three point divided difference method, with the rst 7 s of data neglected to remove transient effects. The wall shear rate was calculated as the ow rate, Q, divided by the cubed height of the channel, DH. Building upon the relation of the half life to Reynolds number, a fourth order curve t demonstrates that the decay of ferrouid is non-linearly proportional to the wall shear rate, as observed by the varying temporal washaway of the aggregate. Cases with higher applied magnetic elds are plotted with open symbols while those with lower elds are shown with lled symbols. The measured rate of dispersion versus wall shear rate can be compared with the 1993 work of Ruuge and Rusetski,19 which is also plotted on Figure 15 as a dashed line. As extracted from the work of Ruuge and Rusetski, the decay rate is signicantly lower than the decay rates from data obtained from this experiment. One source of this difference is the high value of FM from Ruuge and Rusetski, of 10 600
FIG. 15. The time-averaged change in aggregate size plotted versus wall shear rate. The solid linear curve t developed from the data corresponding to the low (lled symbols) and high magnetic elds (open symbols). The dashed line corresponds to a t of the rate of change versus wall shear rate measured in Ruuge and Rusetski.19
dynes/cm3. Conversely, the FM corresponding to the present dispersion data are on the order of 10002000 dynes/cm3 at the aggregate interface. Therefore, when comparing the two expected decay lines, the decay rate from Ruuge and Rusetski should be slower, as is the case in Figure 15 due to the higher magnetic body forces. An additional result from this analysis, intuitively, is that both the rate of decay and aggregate half lives are related to the applied magnetic eld as well as incident ow. However, the nature of the average derivative is divergent between the data presented in Ruuge and Rusetski and the calculated aggregate derivatives measured herein. Given the difference in the decay rates between the two works, a comparison of the two sets of data from applied magnetic elds from the present work can be explored. It is probable that since the interfacial interaction changes signicantly between the two magnetic elds presented herein, the regimes studied in this work may represent a transitional region for aggregate dynamics, particularly given the nonlinearity associated with the time-averaged derivative shown in Figure 15. As the magnetic elds are changed to higher or lower values, the aggregate would be expected to be either fully retained or immediately dispersed when exposed to an incident ow. As an additional distinction, Ruuge and Rusetski did not note any unsteady shedding behavior in their experiment, leading to the conclusion that washaway of their ferrouid aggregates fell into a different regime. One explanation is associated with their selection of magnetic uid, given that they examine saturation magnetization uids a maximum of 1.4 gauss, whereas the EMG 705 uid has a saturation magnetization of 220 gauss, which would provide signicantly higher resistance to ow pressure. Moreover, the volume of injected ferrouid in the work of Ruuge and Rusetski was between 0.05 to 0.3 ml within a 3 to 12 mm channel diameter. In the data presented herein, 2 ml of ferrouid was injected for all cases within a 15 mm hydraulic diameter. Following the half-life and decay rate, we will next discuss a parameter that synthesizes the major forces acting on the aggregate and capture its dynamics, expanding beyond these parameters. Based on the data collected, we infer that the dispersion process is governed by the interaction of the inertial and viscous forces and the balance between magnetic stresses and pressure on the aggregate-ow interface. Since the aggregate behaves as a compliant boundary, it creates a blockage, forcing localized ow acceleration. The presence of the magnetic eld counters ow pressures, acting to retain the aggregate in the region of high eld gradients. However, the interfacial dynamics between magnetic forces and ow shear creates an apparent viscous boundary that gives rise to instabilities, evolving into the Kelvin-Helmholtz shedding process. To further explore this interplay, the magnetic body force per specic volume was extracted at the half aggregate size, evaluated at the aggregate-bulk ow interface, using Eq. (3). A denominator that non-dimensionalizes the magnetic forces is given by M lfshedding;50% ; h50% (5)
127102-10
Phys. Fluids 23, 127102 (2011)
where l refers to the dynamic viscosity of water at ambient temperature, 0.01 poise. The parameter fshedding, 50% is the shedding frequency measured at the time of 50% aggregate size, converted to radians per second, and nally, h50% is the aggregate height at the 50% time in centimeters. Here, a distinction arises between the normalization shown in Figure 12 in that the non-dimensionalization is performed at the aggregate half-life. When studying the relative effects of the magnetic body forces and ow pressures, the half-life is the most representative means to account for the dynamic behavior of the aggregate. Using Eq. (5) to non-dimensionalize the magnetic forces results in the parameter F* F M50% rH50% : M (6)
measurement of the aggregate, while the aggregate is subjected to three-dimensional effects. An additional limitation of this work is the assessment of the magnetic eld effect based on datasets containing two magnetic elds. To fully generalize the non-dimensionalization, additional investigations must be undertaken. Despite these limitations, our results expand previous studies of ferrouid aggregates. For the rst time, we characterize the development of the Kelvin-Helmholtz instability by measuring the shedding frequency of the ferrouid. Also, the half life time model proposed by Ganguly et al. has been extended and corrected to account for a larger span of ows as well as different magnetic elds. Finally, the parameter F* has been developed accounting for the inuence of ow and magnetic elds.
IV. CONCLUSIONS
The results of applying Eq. (6) are shown in Figure 16. Data from both magnetic elds is included, but is limited to cases with a dened half-life. Examining the parameters of Eq. (6) more closely, the values of M50% increase relatively linearly as a function of Re. On the other hand, the denominator of F* increases as U2 and is related to an unsteady ow pressure. Therefore, the physical extrapolation of F* is that even though the height of the aggregate is lower, in the cases of the higher Reynolds number, there is proportionally more aggregate shedding, even while the magnetic body forces are increasing. This competition between the parameters, captured in the F* equation, results in an overall decay curve which goes as the Reynolds number and is accurately described by an exponential curve t with a linearized R2 value of 0.98. Each of the three parameters provides insight into the generalized mechanics of ferrouid aggregates. The relative lengthening of the aggregate during increases of Re discussed in preceding paragraphs results in increased drag forces. The drag maintains aggregate shedding and also restricts the ability of the magnetic body forces to retain the aggregate, even as the aggregate height decreases. Though this work has expanded the body of knowledge related to ferrohydrodynamics, one limitation of this analysis is the reliance on a two-dimensional
The behavior of ferrouid aggregates was studied herein for steady ow conditions over a range of Reynolds numbers spanning 100 to 1000. Three regimes apparent for ows of Re between 100 and 1000 were identied based on maximum magnetic elds of 2 105 and 2.4 105 A/m. Reynolds numbers less than 300 for both elds result in a stable aggregate that stretches in the x and y-directions by up to 8%. At the transitional Reynolds numbers, the aggregate stretches signicantly, losing coherence as it shears away from high magnetic eld gradients. Ferrouid stretching leads to interfacial ripples and light ferrouid tendrils shed periodically downstream. Further increase of the Reynolds number results in Kelvin-Helmholtz shedding from the main aggregate and aggregate erosion of up to 80 percent over 60 s. Generally, the temporally sampled Strouhal number increases with Reynolds number, since the increased Re drives increases in shedding frequency. The inuence of different magnetic elds results in aggregate shedding initiation at a lower Reynolds number as well as an increase in the shedding frequency for a given Reynolds number in the shedding regime. The aggregate half life, decay rate, and a newly developed dimensionless body force parameter are used to holistically elucidate aggregate dynamics. The aggregate half life decreases proportionally to Re2.2, multiplied by an additional correction factor to account for magnetic body forces. In addition, the decay rate of ferrouid is proportional to the wall shear rate. Finally, a dimensionless parameter containing the specic volume-based magnetic body force, along with the aggregate height and shedding frequency, calculated at the aggregate half-life captures the data effectively. A resulting empirical t is proportional to Re5.3, indicating a power law decay relationship between the interaction of aggregatedirected forces and the bulk ow.
1
FIG. 16. Plot of the normalized magnetic body force per unit volume versus the Reynolds number. The results of a power law curve t are shown using the black dashed line.
J. C. Bacri, R. Perzynski, M. I. Shliomis, and G. I. Burde, Negativeviscosity effect in a magnetic uid, Phys. Rev. Lett. 75, 2128 (1995). M. I. Shliomis, Comment on Structure of ferrouid dynamics, Phys. Rev. E 67, 043201 (2003). 3 B. Abou, J. E. Wesfreid, and S. Roux, The normal eld instability in ferrouids: Hexagon-square transition mechanism and wavenumber selection, J. Fluid Mech. 416, 217 (2000). 4 A. G. Boudouvis, J. L. Puchalla, L. E. Scriven, and R. E. Rosensweig, Normal eld instability and patterns in pools of ferrouid, J. Magn. Magn. Mater. 65, 307 (1987).
2
127102-11
5

17
Phys. Fluids 23, 127102 (2011) A. S. Lubbe, C. Bergemann, H. Riess, F. Schriever, P. Reichardt, K. Possinger, M. Matthias, B. Dorken, F. Herrmann, R. Gurtler, P. Hohenberger, N. Haas, R. Sohr, B. Sander, A.-J. Lemke, D. Ohlendorf, W. Huhnt, and D. Huhn, Clinical experiences with magnetic drug targeting: A phase I study with 4-epidoxorubicin in 14 patients with advanced solid tumors, Cancer Res. 56, 4686 (1996). 18 A. N. Rusetski and E. K. Ruuge, Magnetic uid as a possible drug carrier for thrombosis treatment, J. Magn. Magn. Mater. 85, 299 (1990). 19 E. K. Ruuge and A. N. Rusetski, Magnetic uids as drug carriers: Targeted transport of drugs by a magnetic eld, J. Magn. Magn. Mater. 122, 335 (1993). 20 R. Ganguly, A. P. Gaind, S. Sen, and I. K. Puri, Analyzing ferrouid transport for magnetic drug targeting, J. Magn. Magn. Mater. 289, 331 (2005). 21 R. Ganguly and I. K. Puri, in Advances in Applied Mechanics, edited by H. A. a. E. v. d. Giessen (Elsevier, Orlando, Florida, 2007), Vol. 41. 22 R. Ganguly, B. Zellmer, and I. K. Puri, Field-induced self-assembled ferrouid aggregation in pulsatile ow, Phys. Fluids 17, 097104 (2005). 23 N. Otsu, Threshold selection method from grayscale histograms, IEEE Trans. Syst. Man Cybern. 9, 62 (1979). 24 V. Madisetti and D. B. Williams, The Digital Signal Processing Handbook (CRC, New York, 1998). 25 S. L. Jae and V. O. Alan eds, Advanced Topics in Signal Processing (Prentice-Hall, Inc., Englewood Cliffs, NJ, 1987). 26 A. Roshko, Experiments on the ow past a circular cylinder at very high Reynolds number, J. Fluid Mech. 10, 345 (1961). 27 C. H. K. Williamson, Oblique and parallel modes of vortex shedding in the wake of a circular cylinder at low reynolds numbers, J. Fluid Mech. 206, 579 (1989). 28 M. Yokoyama and O. Mochizuki, Deformation of a uid-lled compliant cylinder in a uniform ow, J. Fluids Struct. 25, 1049 (2009). 29 T. Weser and K. Stierstadt, Magnetoviscosity of concentrated ferrouids, Z. Phys. B 59, 257 (1985). 30 R. E. Rosensweig, Ferrohydrodynamics (Cambridge University Press, New York, 1985), republished by Dover, 1997.
M. Zahn and R. Rosensweig, Stability of magnetic uid penetration through a porous medium with uniform magnetic eld oblique to the interface, IEEE Trans. Magn. 16, 275 (1980). 6 G. Pacitto, C. Flament, J. C. Bacri, and M. Widom, Rayleigh-Taylor instability with magnetic uids: Experiment and theory, Phys. Rev. E 62, 7941 (2000). 7 S. K. Malik and M. Singh, Finite amplitude KelvinHelmholtz instability in magnetic uids, Phys. Fluids 29, 2853 (1986). 8 G. Moatimid, Nonlinear Kelvin-Helmholtz instability of two miscible ferrouids in porous media, ZAMP 57, 133 (2005). 9 J. A. Miranda and M. Widom, Parallel ow in Hele-Shaw cells with ferrouids, Phys. Rev. E 61, 2114 (2000). 10 K. Raj and R. Moskowitz, Commercial applications of ferrouids, J. Magn. Magn. Mater. 85, 233 (1990). 11 C. Alexiou, W. Arnold, P. Hulin, R. J. Klein, H. Renz, F. G. Parak, C. Bergemann, and A. S. Lubbe, Magnetic mitoxantrone nanoparticle detection by histology, x-ray and MRI after magnetic tumor targeting, J. Magn. Magn. Mater. 225, 187 (2001). 12 C. Alexiou, W. Arnold, R. J. Klein, F. G. Parak, P. Hulin, C. Bergemann, W. Erhardt, S. Wagenpfeil, and A. S. Lubbe, Locoregional cancer treatment with magnetic drug targeting, Cancer Res. 60, 6641 (2000). 13 C. Alexiou, R. Jurgons, R. Schmid, A. Hilpert, C. Bergemann, F. Parak, and H. Iro, in vitro and in vivo investigations of targeted chemotherapy with magnetic nanoparticles, J. Magn. Magn. Mater. 293, 389 (2005). 14 C. Alexiou, R. Jurgons, R. J. Schmid, C. Bergemann, J. Henke, W. Erhardt, E. Huenges, and F. Parak, Magnetic drug targetingbiodistribution of the magnetic carrier and the chemotherapeutic agent mitoxantrone after locoregional cancer treatment, J. Drug Target. 11, 139 (2003). 15 C. Alexiou, A. Schmidt, R. Klein, P. Hulin, C. Bergemann, and W. Arnold, Magnetic drug targeting: Biodistribution and dependency on magnetic eld strength, J. Magn. Magn. Mater. 252, 363 (2002). 16 A. S. Lubbe, C. Alexiou, and C. Bergemann, Clinical applications of magnetic drug targeting, J. Surg. Res. 95, 200 (2001).

Dispersion of Ferrofluid Aggregates in Steady Flows: Related Articles

Uploaded by

Copyright:

Available Formats

Dispersion of Ferrofluid Aggregates in Steady Flows: Related Articles

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dispersion of Ferrofluid Aggregates in Steady Flows: Related Articles

Uploaded by

Copyright:

Available Formats

Dispersion of ferrofluid aggregates in steady flows

Additional information on Phys. Fluids

PHYSICS OF FLUIDS 23, 127102 (2011)